Research On Falling Alarm System Based On Parallel Detection Of Characteristic Parameter Changing Rate

With the development of an aging society,the guardianship of the safety and health of elderly people living alone has become a growing concern of most middle-aged people.Sudden falls often bring both psychological and physical blows to empty-nesters and patients in rehabilitation.The secondary damage caused by failing to be found and rescued in time after falls is a huge burden on society and the family.Considering the shortcomings of the current fall alarm systems,such as high cost,inconvenient to carry,applicable to single application scenario,and difficult to popularize in the market on a large scale,we design a wearable fall detection and alarm system with mulitiple application scenarios by combining the single chip computer Arduino and Android system,without affecting the daily life of the guardian,providing timely and various ways to alarm and fall detection anytime and anywhere.Based on the analysis of the key technologies and related research in the field of fall detection,we develop in this thesis a fall detection system which integrates six-axis inertial motion sensor Mpu6050 data acquisition,single-chip Arduino data processing and algorithm implementation,and an APP called "Helper" under Android system.The main research work of this paper is summarized as follows:(1)With analysis of the advantages and disadvantages of the existing technology research and system,the overall system framework and scheme are determined,the hardware circuit of the system is designed and implemented,the working principles of data acquisition module Mpu6050 and Bluetooth module HC05 are analyzed,and the single-chip Arduino Nano is selected as the core processing module to implement the related algorithms,and the APP--"Helper"in Android system is designed and implemented as the core part of positioning and alarming.In addition,the thesis mainly introduces the principle of location positioning,the realization logic of sending short messages with a certain frequency and number of times,and a one-click call function,which makes the transmission mode of alarm information more timely,effective and diversified.(2)After further categorizing the daily motion scenes of the human body and performing human body modeling to determine the sampling frequency and the wearing position of the device,we choose the collected characteristic values to include the amplitude of the tri-axial acceleration of the body(AM),and the amplitude of the combined acceleration relative to the rate of change before a step period(*AM),and the angle change between the trunk direction and the ideal vertical direction(AG).The measurement of the*AM more intuitively reflects the degree of dramatic changes in the monitored object relative to normal motion.(3)Since the human body moves periodically with the pace,the range of AG when moving up and down stairs is larger than that when body walks.Misjudgment for falls on the way up and down stairs may occur,because the change of AG is less than the dangerous threshold and the intensity of movement is less than the walking fall.Through the difference of the periodic fluctuation range of the AG,the movement state of daily walking and moving up and down stairs are accurately distinguished.Based on the simulation experiments,the danger thresholds of the two scenarios are determined,respectively.It improves the detection accuracy and expands the application scenarios of the system.(4)With the analysis of the characteristic values of six common movements and two types of fall behaviors,a parallel threshold detection algorithm based on the changing rate of characteristic parameters is proposed,which combines sensor information fusion and the Protothread model.In this thesis,we compare the experimental results with the classical threshold algorithm and SVM algorithm by simulating the behavior of the elderly,and verify the accuracy and reliability of the algorithm.